大梁自动焊起始点定位及障碍物识别的关键技术研究

发布时间：2018-10-22 09:53

【摘要】：随着中国“智能制造”创新战略的深入推进,传统制造业的转型升级面临巨大的机遇与挑战。焊接作为传统制造业中不可或缺的材料加工方法,实现焊接智能创新升级及全自动化已成为当前行业发展的必然趋势。大梁是现今重型装备行业中最普遍的焊接结构件形式之一,广泛应用于桥梁、轨道交通、运输集装箱、起重机械、房屋建筑等领域。目前,由于大梁工件在焊接生产线的变位机械装夹或中心翻转过程中,存在由人工操作引起的垂直方向高度及偏转角度随机误差,难以实现焊缝起始点自主寻位;同时,大梁工件上随机分布流水槽、加强板、辅助上翼板等障碍物,具有尺寸多变性、结构多元化等特点,难以通过传统的视觉识别算法实现大梁工件的障碍物识别聚类及智能规避。本文针对大梁焊接全自动化尚存的几大技术难点,对传统半自动化大梁生产线进行智能改造和创新升级,深入研究大梁自动焊起始点定位及障碍物识别的关键技术。提出一种复合检测式双涡流定位传感的焊接起始点自动定位方法。依据涡流传感器的影响因素及规律,阐述涡流检测高度及面积的机理。建立双涡流定位系统模型,通过电感计算及等效分析,探究得到复阻抗的变化规律。搭建简易定位数据采集试验平台,优化采样策略,获取双探头输出特征值。采用加权最小二乘法拟合探头1输出特征值与高度函数,基于响应面法分离变量,代入探头1高度值,得出探头2输出特征值与偏转角度的拟合函数。根据复合检测式双涡流定位传感器焊接起始点定位方案及三个主要定位特征参数,计算焊枪横纵方向距离偏差值。提出一种折线式线阵激光传感器障碍物实时聚类识别方法。根据三角法原理,设计折线式激光传感器光路平面布局及光路系统参数。基于dsPIC30f4012单片机实现激光传感器硬件电路设计。简要分析激光视觉传感器信号采集成像机理,设计线激光有效长度自适应统一化优化方案确保障碍物识别稳定可靠。针对传统FCM应用于大梁障碍物实时聚类识别过程中存在的缺陷,引入实时聚类策略,替换距离函数,全局快速优化,得到一种优化模糊C均值实时聚类算法。通过MATLAB平台仿真对比分析,证明该算法能实时获取聚类数及障碍物类型属性。在大梁自动焊模拟平台进行焊接起始点定位试验,结果表明动态响应快,适用范围广,横向及纵向定位精度高。在某公司大梁自动焊生产线平台进行障碍物实时识别试验,结果表明实时性优良,聚类数与实际相符,障碍物规避动作精准。
[Abstract]:With the development of innovation strategy of intelligent manufacturing in China, the transformation and upgrading of traditional manufacturing industry is faced with great opportunities and challenges. Welding, as an indispensable material processing method in traditional manufacturing industry, has become an inevitable trend of current industry development. Beam is one of the most common welding structures in heavy equipment industry. It is widely used in bridge, rail transit, transportation container, lifting machinery, housing construction and so on. At present, due to the random errors of vertical height and deflection angle caused by manual operation, it is difficult to realize the self-locating of the starting point of the weld due to the random errors of vertical height and deflection angle caused by manual operation in the process of mechanical clamping or center flipping of the beam workpiece in the welding production line; at the same time, The obstacles such as randomly distributed flow flume, stiffening plate and auxiliary upper wing plate on the workpiece of beam have the characteristics of variable size and diversified structure, so it is difficult to realize obstacle identification clustering and intelligent evasion by traditional visual recognition algorithm. In this paper, aiming at the remaining technical difficulties of full automation of beam welding, the traditional semi-automatic beam production line is innovated and upgraded intelligently, and the key technologies of locating the starting point of automatic welding and identifying obstacles are deeply studied. This paper presents an automatic welding starting point location method based on compound detecting dual eddy current positioning sensor. According to the influencing factors and rules of eddy current sensor, the mechanism of eddy current measuring height and area is expounded. The model of dual eddy current positioning system is established, and the variation law of complex impedance is obtained by inductance calculation and equivalent analysis. Set up a simple location data acquisition test platform, optimize the sampling strategy, and obtain the output eigenvalue of double probe. The weighted least square method is used to fit the eigenvalue and height function of probe 1. Based on the method of response surface, the fitting function of eigenvalue and deflection angle of probe 2 is obtained by separating variables and replacing the height of probe 1 with the method of response surface. According to the location scheme of welding starting point and three main positioning characteristic parameters of the compound detecting dual-eddy current positioning sensor, the deviation value of the distance between transverse and longitudinal direction of the welding torch is calculated. A real-time clustering method for obstacle recognition of linear laser sensor is presented. According to the triangulation principle, the optical path layout and optical path system parameters of the broken line laser sensor are designed. The hardware circuit of laser sensor is designed based on dsPIC30f4012 single chip computer. The principle of laser vision sensor signal acquisition and imaging is analyzed briefly, and an adaptive unified optimization scheme of line laser effective length is designed to ensure the stability and reliability of obstacle identification. Aiming at the defects of traditional FCM application in the real-time clustering identification of beam obstacles, a fuzzy C-means real-time clustering algorithm is obtained by introducing the real-time clustering strategy, replacing the distance function, and rapidly optimizing the whole world. The simulation results of MATLAB platform show that the algorithm can obtain clustering number and obstacle type attributes in real time. The results show that the dynamic response is fast, the range of application is wide, and the accuracy of transverse and longitudinal positioning is high. A real-time obstacle identification test was carried out on the platform of a company's automatic beam welding production line. The results show that the real-time performance is good, the clustering number is consistent with the actual situation, and the obstacle avoidance action is accurate.
【学位授予单位】：湘潭大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TG409

【参考文献】